Biochimica et Biophysica Acta

Transcription

1 Biochimica et Biophysica Acta 1828 (2013) Contents lists available at SciVerse ScienceDirect Biochimica et Biophysica Acta journal homepage: Investigation of interfacial behavior of glycyrrhizin with a lipid raft model via a Langmuir monolayer study Seiichi Sakamoto a, iromichi Nakahara a, Takuhiro Uto b, Yukihiro Shoyama b, samu Shibata a, a Department of Biophysical Chemistry, Faculty of Pharmaceutical Sciences, Nagasaki International University, uis Ten Bosch, Sasebo, Nagasaki , Japan b Department of Pharmacognosy, Faculty of Pharmaceutical Sciences, Nagasaki International University, uis Ten Bosch, Sasebo, Nagasaki , Japan article info abstract Article history: Received 31 July 2012 Received in revised form 8 January 2013 Accepted 9 January 2013 Available online 16 January 2013 Keywords: Glycyrrhizin Langmuir monolayer Lipid raft Membrane disruption Saponin An interaction of glycyrrhizin (GC) with a lipid raft biomembrane model that consisted of N-palmitoyl- D-erythro-sphingosylphosphorylcholine (PSM), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DPC), and cholesterol (CL) was systematically studied using the Langmuir monolayer technique. To construct the lipid raft model, the surface pressure (π) molecular area (A) and surface potential (ΔV) A isotherms for three-component (PSM/ DPC/CL) systems on 0.02 M Tris buffer with 0.13 M NaCl (p 7.4) were primarily measured by changing their compositions. Thermodynamic and interaction parameters for binary PSM/DPC and PSM/CL systems revealed that PSM interacts more strongly with CL than with DPC. In addition, a morphological analysis performed with Brewster angle microscopy (BAM) and fluorescence microscopy (FM) revealed an optimal ratio of PSM/DPC/CL (1/1/1, by mole) as a model of lipid rafts. Second, the interaction of GC with the ternary PSM/ DPC/CL monolayers was investigated on Tris buffer solutions containing different GC concentrations (1, 5, 10, 25, and 50 μm). In BAM and FM images, microdomains were found to become smaller by increasing the GC concentration in the subphase, suggesting that GC regulates the size of raft domains, which provide dynamic scaffolding for numerous cellular processes. More interestingly, the distinctive GC striped regions were formed at the interface at 50 μm, which shows that GC divides the ternary monolayer into pieces. This phenomenon was observed only in the presence of CL in the monolayer. These results suggest that CL plays an essential role in the interaction with GC, which results in one of the major activities associated with saponins' membrane disruption Elsevier B.V. All rights reserved. 1. Introduction Recently, the demand for Japanese traditional herbal medicines, the so-called Kampo medicines, has been dramatically increasing due to the rising concerns with health foods and complementary and alternative medicines [1]. The root of licorice belonging to the genus Glycyrrhiza, including Glycyrrhiza uralensis and Glycyrrhiza glabra (Chinese licorice), is the most widely used natural resource in Japanese traditional herbal medicines and is included in more than 70% of prescribed Kampo medicines. Furthermore, licorice has also been extensively utilized worldwide as a sweetener, tobacco flavoring agent, and food additive, and in cosmetics and confectionery foods [2]; accordingly, it has been recently named Medicinal Plant of the Year A major bioactive compound of licorice, glycyrrhizin (GC; Fig. 1) has been shown to exhibit numerous pharmacological activities such as hepatoprotective [3], anti-inflammatory [4,5], anti-cariogenic [6], and anti-tumor activities [7]. In addition, various antiviral activities against coronavirus associated with severe acute respiratory syndrome [8], Newcastle disease virus, herpes simplex Corresponding author. Tel./fax: address: wosamu@niu.ac.jp (. Shibata). URL: (. Shibata). virus type 1, vesicular stomatitis virus [9], human immunodeficiency virus type 1 [10,11], influenza A virus [12], and hepatitis A virus [13] have been extensively studied. Due to these outstanding activities, GC has been clinically used for the treatment of allergies and chronic viral hepatitis for more than 50 years [14]. Moreover, GC has attracted much attention as a food additive, especially in natural sweetener with a low calorific value, because it has been shown to exhibit a sweetness 50-times more potent than that of sucrose [15]. Physicochemical properties of 18α- (trans-configuration at the C-18 position) and 18β-GC (cis-configuration at C-18 position) have been investigated by Kondo et al. [16]. These have been found to display interesting properties, particularly with respect to surface activity [17 19] and gel formation [20]. Because of the structural characteristics of amphiphiles, both exhibit interfacial activities with a critical micelle concentration (CMC) of ca. 0.2 mm, which was determined by measuring the interfacial tension at water liquid paraffin interface at 25 C. A notable dissimilarity between the physicochemical properties of the two isomers is related to their abilities to form gels. The aqueous solution of 18β-GC forms an extremely rigid gel even at a dilute concentration of less than 0.1 wt% (1.2 mm) at below p 4.5, whereas 18α-GC does not. Microdomains rich in sphingomyelin (SM) and cholesterol (CL), which are commonly referred to as lipid rafts, have been a subject of /$ see front matter 2013 Elsevier B.V. All rights reserved.

2 1272 S. Sakamoto et al. / Biochimica et Biophysica Acta 1828 (2013) A C C C Glycyrrhizin (GC) B N P - N + N-palmitoyl-D-erythro-sphingosylphosphorylcholine (PSM) P - N + 1,2-dioleoyl-sn-glycero-3-phosphocholine (DPC) Cholesterol (CL) Fig. 1. Structures of (A) glycyrrhizin (GC) and (B) N-palmitoyl-D-erythro-sphingosylphosphorylcholine (PSM), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DPC), and cholesterol (CL). great interest in cell biology because the significant membrane proteins involved in their physiological functions may be located selectively within them. Accordingly, they constantly provide dynamic scaffolding for various different cellular processes, including protein trafficking, signal transduction, and transportation of CL and membrane [21,22]. In rafts, long saturated acyl chains in sphingolipids (both SM and glycosphingolipids) are believed to tightly interact and to be packed with CL, which results in organization of liquid-ordered (l o ) phases [23 26]. By contrast, unsaturated phospholipids are loosely packed to form liquid-disordered (l d ) phases. These different packing properties lead to phase separation [27]. To elucidate liquid liquid immiscibility in membranes, a large number of physicochemical studies on lipid rafts have been performed using a lipid mixture that mimics components of the outer leaflet of plasma membranes, in which the combination of sphingolipids, phospholipids, and CL has been most widely utilized. Recently, the monolayer technique, in combination with the Langmuir Blodgett method [28], atomic force microscopy (AFM) [29 32], Brewster angle microscopy (BAM) [33,34], and various kinds of fluorescence microscopy (FM) [35 41], has been extensively performed using raft-mimicking lipid mixtures in order to examine their lateral organization. These methods enable a significant amount of important information concerning phase variations on micro- and nano-scales at the air/water interface to be obtained. The behavior of GC on plasma membranes was evaluated by Nakamura et al. using cultured rat hepatocytes treated with hepatotoxin [42]. Release of glutamic-oxaloacetic transaminase and lactic dehydrogenase, which is induced by hepatotoxin, is significantly reduced by increasing the concentration of GC. These results suggest that GC directly prevents enzyme leakage by changing membrane permeability of hepatocytes. More recently, arada reported the anti-viral strategy of GC and showed that the inhibition of viral infection by GC is mainly due to the suppression of fluidity of the plasma membrane and viral envelope into which GC is incorporated [43]. Interestingly, in both cases, GC acts to lower fluidity of the plasma membrane. owever,

3 S. Sakamoto et al. / Biochimica et Biophysica Acta 1828 (2013) morphological changes related to the ability of GC to suppress membrane fluidity remain unknown. GC is structurally classified as a triterpenoid saponin that possesses both hydrophobic (triterpenoid skeleton) and hydrophilic moieties (glucuronic acid) in the molecule. Saponins have been shown to cause membrane perturbation, which is frequently referred to as the hemolytic activity of saponins [44 48]. The molecular mechanism for the membrane-permeabilizing activity of saponins and steroidal glycoalkaloids has been proposed [49,50]. Briefly, saponins integrate with their hydrophobic part to form a complex with sterols, and formation of this complex promotes interaction of their extra-membranous orientated saccharide moieties, which results in their accumulation into plaque. Steric interference of these saccharide moieties causes membrane curvature, which subsequently leads to pore formation [49] or vesiculation [50] in the membrane. Alternatively, after their integration into the membrane, saponins may migrate toward sphingolipid/sterol-enriched membrane domains before forming a complex with the incorporated sterols, thereby interfering with specific domain functionalities [51]. owever,morphological observation that strongly supports this hypothesis has not yet been reported. erein, we have investigated the effect of GC on a lipid raft model by employing the Langmuir monolayer technique. As representative components of lipid rafts in the model membrane, N-palmitoyl-Derythro-sphingosylphosphorylcholine (PSM), 1,2-dioleoyl-sn-glycero-3- phosphocholine (DPC), and CL (Fig. 1) were used. In the present study, we focused on the interaction of GC adsorbed from the subphase solutions with the raft model monolayer. The surface pressure (π) molecular area (A) and surface potential (ΔV) A for single, binary, and ternary monolayers composed of PSM, DPC, and CL in the absence and the presence of GC in the subphase were primarily measured under physiological p and ionic-strength conditions. Subsequently, GC dose-dependent phase behavior and domain variation of the model membrane were morphologically revealed using BAM and FM. 2. Materials and methods 2.1. Materials N-palmitoyl-D-erythro-sphingosylphosphorylcholine (PSM; purity> 99%), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DPC; purity > 99%), and 1-palmitoyl-2-{6-[(7-nitro-2-1,3-benzoxadiazol-4-yl)amino] hexanoyl}-sn-glycero-3-phosphocholine (NBD-PC; purity>99%) were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL, USA). Cholesterol (CL; purity 99%) was obtained from Sigma Chemical. Co. (St. Louis, M, USA). These lipids were used without further purification or characterization. Chloroform (99.7%) and methanol (99.8%), used as spreading solvents, were purchased from Kanto Chemical Co., Inc. (Tokyo, Japan) and nacalai tesque (Kyoto, Japan), respectively. Tris(hydroxymethyl) aminomethane (Tris), sodium chloride (NaCl), and acetic acid of guaranteed reagent grade for the preparation of a subphase were obtained from nacalai tesque (Kyoto, Japan). To exclude all surfaceactive organic impurities, NaCl was roasted at 1023 K for 24 h before use. The substrate solution was prepared using thrice distilled water (surface tension=72.0 mn m 1 at K; electrical resistivity= 18 MΩ cm) Purification of glycyrrhizin (GC) The compound 18β-GC (98%) obtained from nacalai tesque (Kyoto, Japan) required purification due to the presence of an impurity band in the thin-layer chromatography analysis. Further purification of GC was carried out by reverse-phase column chromatography using CSMSIL (nacalai tesque, Kyoto, Japan). GC (5.0 g) was dissolved in 10 ml of 70% (v/v) methanol and applied to CSMSIL gel (5.0 cm 55.0 cm) equilibrated with 70% methanol. GC was then eluted with 3.0 L of the mobile phase (70% methanol) to give 4.5 g of purified GC. This purification was repeated seven times to obtain approximately 30 g of purified GC. The purity of GC was then measured by high-performance liquid chromatography analysis using two pumps (DP-8020, Tosoh Corp., Tokyo, Japan), connected with a UV detector (UV-8020, Tosoh Corp., Tokyo, Japan, 254 nm), degasser (SD-8022, Tosoh Corp., Tokyo, Japan), and an RP-18 column (TSK-GEL DS-100 V, 250 mm 4.6 mm, 5 μm, Tosoh Corp., Tokyo, Japan). The mobile phase consisted of 40% acetonitrile that contained 2% acetic acid was used at a flow rate of 0.7 ml min 1. The purity of GC was estimated to be 98% according to the peak-area percentage method applied to the obtained chromatogram Surface pressure area isotherms The surface pressure (π) of monolayers was measured using an automated homemade Wilhelmy balance. The resolution of the surface pressure balance (Mettler Toledo, AG-245) was 0.01 mn m 1. The pressure-measuring system was equipped with filter paper (Whatman 541, periphery=4 cm). The trough was made from Teflon-coated brass (area=750 cm 2 ), and both hydrophobic and lipophobic Teflon barriers were used in this study. Surface pressure (π) molecular area (A) isotherms were recorded at 298.2±0.1 K. Stock solutions of PSM (0.5 mm), DPC (0.5 mm), and CL (1.0 mm) were prepared in chloroform/methanol (2/1, v/v). Spreading solvents were allowed to evaporate for 15 min before compression. The monolayer was compressed at a speed of ~0.10 nm 2 per molecule (without considering GC molecules) per minute. Standard deviations (SD) for the molecular surface area and surface pressure were ~0.01 nm 2 and ~0.1 mn m 1, respectively. The p of the subphase (0.02 M Tris buffer with 0.13 M NaCl) was adjusted to 7.4 with acetic acid Surface potential area isotherms The surface potential (ΔV) was recorded simultaneously with surface pressure when the monolayer was compressed at the air/water interface. This potential was monitored with an ionizing 241 Am electrode at 1 2 mm above the interface, whereas a reference electrode was dipped into the subphase. An electrometer (Keithley 614) was used to measure surface potential. SD of surface potential was 5 mv Surface tension measurements The surface tension (γ) of surfactant solutions was determined at K by using a drop volume tensiometer (YC-2010, YTS, Japan) [52], which measures the volume of a drop detaching from a capillary with a known diameter. Temperature was maintained constant within ±0.03 K by means of a thermostat. The experimental error in estimation of surface tension was ±0.05 mn m Brewster angle microscopy (BAM) The monolayer was directly visualized by using a Brewster angle microscope (KSV ptrel BAM 300, KSV Instruments, Finland) coupled to a commercially available film balance system (KSV Minitrough, KSV Instruments, Finland). The use of a 20-mW e Ne laser emitting p-polarized light with a wavelength of nm and an instrument that was equipped with a 10 objective lens enabled a lateral resolution of ~2 μm. The angle of the incident beam to the air/water interface was fixed to the Brewster angle (53.1 ) at K. The reflected beam was recorded with a high-grade charge coupled device (CCD) camera (EDkamPro02, ED Imaging Gmb, Germany), and BAM images were digitally stored on a computer hard disk.

4 1274 S. Sakamoto et al. / Biochimica et Biophysica Acta 1828 (2013) Fluorescence microscopy (FM) The film balance system (KSV Minitrough, KSV Instruments, Finland) was mounted onto the stage of an lympus BX51W1 microscope (Tokyo, Japan) equipped with a 100-W mercury lamp (US-1030L), an objective lens (SLMPlan 50, SLMPlanFI 20, working distance= 15 mm), and a 3CCD camera control unit (IKTU51CU, Toshiba, Japan). The samples for FM were prepared by addition of 1 mol% of the phospholipid fluorescent probe, NBD-PC, individually to the stock solutions. The images observed at excitation and emission wavelengths of 460 and 534 nm, respectively, were directly recorded onto the hard disk via an online image processor (DVgate Plus, Sony, Tokyo, Japan) connected to the microscope. Image processing and analysis were conducted using the Adobe Photoshop Elements ver. 9.0 (Adobe System Incorporated, CA, USA) software package. The total amount of ordered domains (dark contrast regions) was evaluated and expressed as a percentage per frame by dividing the respective frames into dark and bright regions Fabrication of GC layers adsorbed from the subphase The Tris buffer solutions (0.02 M) with 0.13 M NaCl (p 7.4) containing proper amounts of GC (1, 5, 10, 25, or 50 μm) were prepared as subphases. After clearing trough and barriers with chloroform and acetone, the subphase was poured into the trough. We then waited until an equilibrium state of GC adsorption from the subphase to surface was achieved. The equilibrium time of adsorption depended on the GC concentration in the subphase. After equilibrium, samples such as PSM, DPC, CL, and their mixtures were spread onto the surface. The isotherm measurements and microscopic observations were made in the same manner as that previously described. Note that this is not the method where highly concentrated GC solutions are injected into the subphase. Solubility of GC in the aqueous solution is considerably low. Therefore, if we applied the injection method, large amounts of GC solutions would need to be injected so that the water level of the subphase would increase substantially. owever, in the method used in the present study, there are errors in base values of π and ΔV; indeed, the isotherm data were obtained under the base of respective GC-containing Tris buffer not just Tris buffer. First, γ and ΔV values at an equilibrium state of GC in the 0.02 M Tris buffer solution with 0.13 M NaCl (p 7.4) were measured to normalize the data obtained in this manner (see Fig. S1 in Supplementary material). Using these values, all π A and ΔV A isotherms shown herein were normalized in subsequent procedures. In the present study, experimental π (=π exp )isobtainedbyeq.(1): π exp ¼ γ GC γ m where γ GC denotes the surface tension of the GC-containing buffer solution at a certain concentration of GC (C GC ) and γ m represents the surface tension of the monolayer on the GC-subphase. The γ GC values were independently measured using the drop volume method (Fig. S1(A)). erein, we define π GC as follows: π GC ¼ γ 0 γ GC where γ 0 is the surface tension of the buffer solution without any surfactants. From Eqs. (1) and (2), π is offset so that the base value of π comes to become γ 0 : π ¼ π exp þ π GC ¼ γ 0 γ m The ΔV value can be normalized similarly. The experimental value (ΔV exp ) is expressed in Eq. (4): ΔV exp ¼ V m V GC ð1þ ð2þ ð3þ ð4þ where V m and V GC denote potentials of the subphase covered with the monolayer and those of the GC-containing buffer solution at a certain C GC, respectively. In advance, surface potential (ΔV GC ) of GC solutions was obtained in the absence of lipid monolayers (Fig. S1(B)). The ΔV GC value satisfies Eq. (5): ΔV GC ¼ V GC V 0 where V 0 represents the potential of the buffer solution in the absence of surfactants. We can obtain the normalized value from Eqs. (4) and (5) as follows: ΔV ¼ ΔV exp þ ΔV GC ¼ V m V 0 That is, a zero of vertical axes of π and ΔV, plotted in the whole figures, means those of a non-covered surface of 0.02 M Tris buffer solution with 0.13 M NaCl (p 7.4) in the absence of lipid monolayers and GC. 3. Results and discussion 3.1. ne-component monolayers for N-palmitoyl-D-erythro-sphingosylphosphorylcholine (PSM), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DPC), and cholesterol (CL) The surface pressure (π) molecular area (A) and surface potential (ΔV) A isotherms of representative PSM, DPC, and CL were measured on 0.02 M Tris buffer with 0.13 M NaCl (p 7.4) at K (Fig. 2). The π A isotherm of PSM shows a phase transition pressure (π eq )of~21 mnm 1 (solid arrow), at which the monolayer phase starts changing from a liquid-expanded (LE) to liquid-condensed (LC). The extrapolated area, which is the molecular area of the ordered-state part on the π A isotherm at π=0, of PSM is ~0.48 nm 2. DPC forms a typical expanded monolayer with an extrapolated area of ~0.79 nm 2.Incontrast, CL forms a typical condensed monolayer (liquid or solid state, see Fig. S2 in Supplementary material) with a limiting molecular area of ~0.38 nm 2, which is in good agreement with the cross-sectional area of CL molecules computed on the basis of average tilt angle [53].Values of π eq, phase collapse pressure (π c ), and its corresponding collapse area (A c ) of PSM, DPC, and CL are shown in Table 1. The surface potential (ΔV) generally reflects orientational and conformational changes of monolayers upon compression. The ΔV values of all monolayers vary positively upon compression, which means that the molecular orientation in a monolayer is improved in the direction normal to the surface plane [54]. When the PSM, DPC, and CL monolayers are compressed up to the close-packed state, their ΔV values reach ~354, ~418, and ~401 mv, respectively. These results are in good agreement with those of previous studies [55 57]. In case of PSM, the slope of ΔV A isotherm changes at ~0.51 nm 2 (indicated by an arrow), which corresponds to the LE/LC phase transition. This result suggests that molecular orientation of PSM is abruptly improved when the phase is changed from the LE to LC state. In situ morphological observations for a one-component monolayer at the air/water interface were carried out by BAM and FM. BAM images of DPC and CL monolayers show a homogeneous LE phase (dark contrast) and an LC phase (bright contrast), respectively (see Fig. S3(B), (C) in Supplementary material). As for PSM, the phase transition (π eq =~21 mn m 1 ) from the LE to LC phase is observed when transition is judged from the dark bright contrast (Fig. S3(A) in Supplementary material). In order to further investigate interfacial behaviors of one-component monolayers, FM possessing greater resolution and magnification was then performed, even though the doping probe was in the sample. Fig. 3 shows fluorescence images of PSM monolayers on 0.02 M Tris buffer (0.13 M NaCl, p 7.4) at K. ð5þ ð6þ

5 S. Sakamoto et al. / Biochimica et Biophysica Acta 1828 (2013) Fig. 2. The π A and ΔV A isotherms of pure PSM, DPC, and CL monolayers spread on 0.02 M Tris buffer containing 0.13 M NaCl at K (p 7.4). Solid arrows in the isotherms indicate an LE/LC phase transition. Fluorescent probes (here, NBD-PC) dissolve selectively into the LE phase in the monolayer state. Therefore, in contrary to bright and dark regions in BAM images, those in FM images correspond to LE and LC phases, respectively. Below π eq =~21 mn m 1, FM images are homogeneously bright. owever, the nucleation of LC domains is apparent at the transition point, and further compression leads to growth of LC domains in starburst-like shapes up to 25 mn m 1. Above 30 mn m 1,edgesof LC domains become unclear. The size of LC domains in PSM remains constant, even though LC domains of 1,2-dipalmitoyl-sn-glycero-3- phosphocholine (DPPC) grow by lateral compression [58].Itisconsiderable that the degree of hydration around the head group of PSM, which possesses amine and hydroxyl groups, is stronger than that of DPPC, which reflects growth of LC domains by lateral compression Binary monolayers (PSM/DPC and PSM/CL) The π A and ΔV A isotherms of two-component (PSM/DPC and PSM/CL) systems were measured on 0.02 M Tris buffer with 0.13 M NaCl (p 7.4) at K (Fig. 4). The isotherms of π A and ΔV A for PSM/DPC system are shown in Fig. 4(A). The inset in Table 1 The phase transition pressure (π eq ), phase collapse pressure (π c ), and its corresponding collapse area (A c ) of PSM, DPC, and CL. π eq (mn m 1 ) π c (mn m 1 ) A c (nm 2 ) PSM DPC CL Fig. 4(A) shows the mean area per molecule (A) of monolayers as a function of X PSM at 25 mn m 1. The dashed line was drawn assuming linear additivity, and solid points represent experimental values. The positive deviation from linearity suggests that PSM repulses DPC by possessing double saturated chain that is considered to prevent packing of DPC monolayers. In addition, the inset of Fig. 4(A) indicates that the π A isotherms shift to smaller areas with increasing the molar fraction of PSM (X PSM ). As to π eq variation, it decreases slightly within a range of ~5.0 mn m 1 as X PSM increases (0.8 X PSM 1), as evident in Figs. S4 and 6(A), whereas π c values gradually vary between the monolayers of DPC (curve 1) and PSM (curve 9) with increasing X PSM (0.7 X PSM 1) (Fig. 6(A)). With respect to ΔV values in this system, no obviously different behavior is observed, except that the ΔV values gradually decrease as X PSM increases. In the binary PSM/CL system (Fig. 4(B)), π A isotherms overlap at 0 X PSM 0.5. In contrast to the binary PSM/DPC system, the additivity of A as a function of X PSM at 25 mn m 1 shows negative deviation, which implies that PSM attracts CL by possessing a large cavity to wrap the CL molecule. In addition, an almost constant mean area per molecule is observed over the range of 0 X PSM 0.5, which results from the overlap observed in the π A isotherms (curves 1 4 infig. 4(B)). Further addition of PSM (X PSM 0.7) expands the isotherm to close to pure PSM monolayers (curve 8) which indicates the LE/LC phase transition at ~21 mn m 1 (solid arrow) at K. The values of π eq slightly decrease as X PSM increases (0.95 X PSM 1), whereas those of π c gradually increase and reach a plateau at X PSM =0.9 (Fig. 6(B)). In ΔV A isotherms, ΔV values increase and reach maxima at X PSM =0.5 and 0.7 (ΔV=~437 mv); then, they decrease to be almost the same as that of the isotherm of pure PSM (curve 8), where the ΔV value is ~354 mv. These results suggest that the PSM/CL monolayer is mostly packed at 0.5 X PSM 0.7. The interaction parameter of PSM with DPC or CL can be evaluated in terms of the excess Gibbs free energy of mixing, ΔG exc. The ΔG exc value is evaluated by integration of π A isotherms from zero to a selected surface pressure on the basis of Eq. (7) [59]: ΔG exc ¼ π ð A 0 12 X 1 A 1 X 2 A 2 Þdπ ð7þ where A i and X i are the area per molecule and molar fraction of the component i, respectively, and A 12 is the mean area per molecule in mixture of components 1 and 2. In absence of interactions between the two-components, ΔG exc =0 [60,61]. The negative and positive deviations from ideality indicate intermolecular attraction and repulsion, respectively. Variations of ΔG exc for the two-component PSM/DPC and PSM/CL monolayers as function of X PSM at five surface pressures are shown in Fig. 5. ΔG exc X PSM curves in the PSM/DPC system at five surface pressures are shown in Fig. 5(A). The ΔG exc values are mostly positive and attain a maximum value at 40 mn m 1 at X PSM =0.6 (800 J mol 1 ). Because their values increase with increasing surface pressure, repulsive interaction between PSM and DPC is induced upon compression. In the PSM/ CL system (Fig. 5(B)), contrary to the PSM/DPC system, the ΔG exc values are negative over the whole range of surface pressures. These results indicate that the interactions between PSM and CL are more attractive than PSM PSM and CL CL interactions. bviously, these values decrease with increasing surface pressures. In particular, they reach ~ 2500 J mol 1 at 0.5 X PSM 0.7 at 45 mn m 1, which indicates that interaction between acyl chains of PSM and the skeleton of CL molecules becomes more attractive as intermolecular distance becomes shorter upon compression. In addition, the ranges (0.5 X PSM 0.7) of the largest ΔG exc values are agreed well with those of the largest ΔV values obtained from ΔV A isotherms of PSM/CL (Fig. 4(B)). These results reflect the fact that lipid rafts organize liquid-ordered (l o ) phases by tightly packing between sphingolipids and CL [23 26]. That is, tight packing between PSM and CL improves the molecular orientation in the

6 1276 S. Sakamoto et al. / Biochimica et Biophysica Acta 1828 (2013) mn m mn m mn m -1 5% 35% 30 mn m mn m µm Fig. 3. FM micrographs of pure PSM monolayers on 0.02 M Tris buffer containing 0.13 M NaCl (p 7.4) at K. The monolayers contain 1 mol% of the fluorescent probe (NBD-PC). The scale bar at the lower-right represents 50 μm. The percentage (%) at the lower-left indicates the ratio of occupied LC domains against the frame. PSM/CL system at 0.5 X PSM 0.7. These results also demonstrate that during monolayer compression, PSM attracts CL and repels DPC. Notably, unsaturated aliphatic chains of DPC monolayers are packed upon compression with π π bonding. Thus, addition of PSM possessing double-saturated chains is considered to prevent monolayer packing of DPC. The interaction between DPC and CL was investigated by the isotherm measurements (π A and ΔV A isotherms) and thermodynamic studies (ΔG exc values)asafunctionofx CL (Fig.S5).Theseresultssuggest that the strongest interaction exists at X CL =0.5 with minimum ΔG exc values of 2000 J mol 1 at 40 mn m 1, which are in good agreement with those in the previous study, indicating that DPC and CL attractively interact each other, especially at X CL =0.5 [62]. owever, a comparison between ΔG exc values in the DPC/CL system ( 1800 J mol 1 at 35 mn m 1 ) and those in the PSM/CL system ( 2300 J mol 1 at 35 mn m 1 ) suggests that CL interacts more strongly with PSM than with DPC through hydrogen bonding. These results support the evidence that the amide nitrogen of sphingomyelin (SM) has a possibility to bind to the 3- group of CL [63,64] for their attractive interaction. Studies of the two-dimensional phase diagram on the transition (π eq ) and the collapse (π c ) pressures in the Langmuir monolayer are Fig. 4. The π A and ΔV A isotherms of the binary (A) PSM/DPC and (B) PSM/CL monolayers on 0.02 M Tris buffer containing 0.13 M NaCl (p 7.4) at K. An inset indicates mean molecular area (A) of the monolayers as a function of X PSM at 25 mn m 1. Solid arrows represent an LE/LC phase transition at ~21 mn m 1.

7 S. Sakamoto et al. / Biochimica et Biophysica Acta 1828 (2013) where x 1 s and x 2 s denote the molar fractions in the two-component monolayers of components 1 and 2, respectively; π 1 c and π 2 c are the collapse pressures of components 1 and 2, respectively; π m c is the collapse pressure of the two-component monolayer at a given composition of x 1 s (or x 2 s ); ω 1 and ω 2 are the corresponding areas per molecule at the collapse points; ξ is the interaction parameter; and kt is the product of the Boltzmann constant and absolute temperature. The solid curve at higher surface pressures was obtained by adjusting the interaction parameter in Eq. (8) to obtain the best fit for the experimental π c values. As shown in Fig. 6, PSM/DPC (0.7 X PSM 1) and PSM/CL (0 X PSM 1) systems show that ξ= 0.40 and ξ= 0.59, respectively. The negative interaction parameters imply higher interaction energy between the molecules relative to the mean energy of interaction between the same molecules. Interaction energy (Δε) is given as follows: Δε ¼ ξrt=z ð9þ where z is the number of nearest neighbors (equal to 6) in a hexagonal close-packed, and the interaction energy is also described as Δε =ε 12 (ε 11 + ε 22 )/2 [67]; ε ij represents the potential energy of interaction between components i and j. The interaction energies were calculated to be 165 J mol 1 (ξ= 0.40) for PSM/DPC system at 0.7 X PSM 1 and 244 J mol 1 (ξ= 0.59) for PSM/CL system. Because the interaction energies are less than 2RT ( J mol 1 ), these two-components are miscible within the range described above Ternary monolayers (PSM/DPC/CL) Fig. 5. Excess Gibbs free energy of mixing for the (A) PSM/DPC and (B) PSM/CL monolayers as a function of X PSM at five different surface pressures. ΔG exc value was calculated from Eq. (7). of much interest since they provide thermodynamic information related to phase behavior. Two-dimensional phase diagrams for binary systems were constructed by plotting π eq and π c values against X PSM at K (Fig. 6). For PSM/DPC (Fig. 6(A)) and PSM/CL (Fig. 6(B)) systems, the π eq change negatively as X PSM increases at 0.8 X PSM 1 (see Fig. S4) and 0.95 X PSM 1, respectively. Regarding the PSM/ DPC system (Fig. 6(A)), the constant π c (0 X PSM 0.7) demonstrates that the two-components are immiscible at 0 X PSM 0.7 although the homogeneous domain (bright regions) is optically observed in FM and BAM images (Fig. S4). A nanometric technique such as AFM may be useful to morphologically investigate miscibility at 0 X PSM 0.7 in greater detail [65,66]. In the present study, however, immiscibility in the range of 0 X PSM 0.7 is clearly described on the basis of thermodynamic considerations. The two-component system is considered to be partially miscible because the PSM/DPC system is miscible at 0.8 X PSM 1. n the other hand, considering the continuous change of π c at 0 X PSM 0.9 and that of π eq at 0.95 X PSM 1, the binary PSM/CL system is miscible in the monolayer state. Assuming that binary monolayers behave as a regular surface mixture with a hexagonal lattice, the coexistence phase boundary between ordered monolayer (2-D phase) and bulk phases (3-D phase) of molecules spread on the surface can be theoretically simulated by the Joos equation [67]: 1 ¼ x s 1 exp π c m π c n s o 2 1 ω1 =kt exp ξ x 2 þ x s 2 exp π c m π c n s o 2 2 ω2 =kt exp ξ x 1 ð8þ The π A and ΔV A isotherms of three-component PSM/DPC/ CL systems are shown in Fig. 7. Since the sphingomyelin (SM) and phosphatidylcholine (PC) have been reported to be contained equally in the lipid raft fraction [68], the ratio of PSM/DPC was fixed in this study. As shown in the figure, the ternary monolayers were investigated as functions of the amount of CL, where an equimolar ratio of PSM/DPC (1/1, mol/mol) was maintained. The π A isotherms appear to shift to small areas and show characteristics of condensed monolayers as the molar fraction of CL increases. In addition, π c values slightly change by the addition of CL. Regarding ΔV A isotherms, maximum ΔV values at the close-packed state increase from ~380 (curve 1) to ~460 mv (curve 6). These results indicate that the addition of CL improves monolayer packing of PSM rather than that of DPC. FM images of the ternary monolayers are shown in Figs. S6 and 8, in which corresponding BAM images are also inserted in the upper-right corner. For the PSM/DPC (1/1) system without CL, both BAM and FM images are optically homogeneous over the whole range of surface pressures, even though microdomains of binary egg SM/DPC (1/1) monolayers have been identified using AFM [31]. Microdomains (so-called raft domains) are observed after addition of a small amount of CL (X CL =0.11), and further addition of CL promotes their formation. Indeed, the percentage at the lower-left in the frame, which represents the ratio of occupied LC domains in the micrograph, increases as the CL amount increases (0.11 X CL 0.5). At 25 mn m 1 (Figs. S6 and 8), the ratio of 84% at X CL =0.5 is greater than 50%. Considering characteristics of PSM (LE/LC) (Fig. 2) and DPC (LE), LC domains are suggested to be composed of CL and PSM. This suggestion is also supported by the ΔG exc analysis, which shows that PSM attractively interacts with CL and repulsively with DPC, as shown in Fig. 5. Therefore, the raft domain of the PSM/DPC/CL (1/1/1) monolayer is found to be CL-rich. Recent studies using PSM/DPC/CL (1/1/1) with the same composition as that used in the present study have reported the round domains, which were assigned to a PSM/CL-rich l o phase [32]. FM images also show that the percentages increase upon lateral

8 1278 S. Sakamoto et al. / Biochimica et Biophysica Acta 1828 (2013) Fig. 6. Two-dimensional phase diagrams based on the variation of the transition pressure (π eq : open circle) and collapse pressure (π c : solid circle) as a function of X PSM ; (A) PSM/ DPC and (B) PSM/CL. The dashed lines were calculated according to Eq. (8). M. indicates a monolayer state. Bulk indicates a solid phase. compression (Figs. S6 and 8). owever, individual raft domains at X CL =0.43 and 0.50 are found to fuse to form larger submicron-sized domains below 35 mn m 1, which corresponds to a lateral surface pressure of biomembrane [69 71]. These fusions are also reproduced in BAM images. n the other hand, at X CL =0.33, the fusion of LC raft domains does not occur until 35 mn m 1. Therefore, the additional ratio of CL is 33% (X CL =0.33) in the equimolar binary PSM/DPC monolayers, and the mixture of PSM/DPC/CL (1/1/1) was selected as a model of lipid rafts in this study. These results support the previously reported data regarding the effect of CL on the size or number of binary egg SM/DPC (1/1) complexes ( nm) [31]. According to the previous report, the size and number of domains are unaffected by the addition of CL to egg SM/DPC (1/1) bilayers in concentrations up to ~15%. When the concentration of CL is 25%, however, the domains become much larger (1 μm in diameter). At 50% CL concentration in egg SM/DPC bilayers, domains are on the order of 10 μm. In addition, the ternary phase diagram, using egg SM/PC/CL, allows us to postulate why high concentrations of CL induce this phenomenon. Egg SM/PC (1/1) mixtures are in the transition region between liquid-disordered (l o ) and solid-ordered phases (s o ). Therefore, the increase in the amount of CL leads to a modification of the domain topology (i.e., formation of l o domains), which causes a large domain (i.e., disappearance of phase separation) at high molar fractions of CL [72] Interaction of PSM/DPC/CL (1/1/1) monolayer with GC Fig. 7. The π A and ΔV A isotherms of the binary PSM/DPC (=1/1, mol/mol) monolayer containing different amounts of CL on 0.02 M Tris buffer with 0.13 M NaCl (p 7.4) at K. (Inset) enlarged ΔV A isotherms in the surface potential region of 300 ΔV (in mv) 500. To elucidate the interaction of GC adsorbed from the subphase with the lipid raft model, which is the equimolar ternary PSM/DPC/CL monolayer, π A and ΔV A measurements were performed. Fig. 9 shows π A and ΔV A isotherms of ternary monolayers on 0.02 M Tris buffer with 0.13 M NaCl (p 7.4) containing different amounts of GC. The lower inserted π time (t) isotherms represent dependence of the

9 S. Sakamoto et al. / Biochimica et Biophysica Acta 1828 (2013) mn m mn m mn m mn m mn m -1 X CL = % 47% 47% 61% 89% X CL = % 60% 75% 93% X CL = % 65% 84% 93% 50 µm Fig. 8. Fluorescence micrographs of the binary PSM/DPC (1/1, mol/mol) monolayer containing different amounts of CL at five surface pressures (5, 15, 25, 35, and 45 mn m 1 )on 0.02 M Tris buffer with 0.13 M NaCl (p 7.4) at K (X CL =0.33, 0.43 and 0.50). The monolayers contain 1 mol% of the fluorescent probe (NBD-PC). The scale bar in the lower-right represents 50 μm. The corresponding BAM images are shown in the upper-right of each photograph. The percentages represent the ratio of LC domains occupied in the frame. adsorption time of GC from the bulk subphase to surface at different GC concentrations (C GC =1,5,10,25,and50 μm). These results indicate that GC attains adsorption equilibrium within 90 min at the previously noted GC concentrations. Therefore, PSM/DPC/CL (1/1/1) monolayers were spread onto the GC solution at 90 min after the subphase containing GC was poured into the trough. The π A isotherms of the PSM/DPC/CL (1/1/1) monolayer are considerably expanded by increasing C GC. owever, it is estimated that these expansions attain saturation at C GC =100 μm since the π A isotherm, obtained from C GC =100 μm, is overlapped with that obtained from C GC =50 μm (data not shown). In the close-packed state, all isotherms have almost the same π c values of ~48 mn m 1 (A=~0.36 nm 2 ). The effect of micelles of GC on PSM/DPC/CL (1/1/1) monolayers was investigated by measuring π A and ΔV A isotherms at C GC levels greater than or equal to critical micelle concentration (CMC) (see Fig. S7 in Supplementary material) [16]. The ΔV A isotherms show that ΔV values on GC subphases finally converge to that without GC upon compression although initial ΔV values are relatively different depending on C GC. As observed in the inserted upper figure, the maximum ΔV value at C GC =10 μm is greater by ~35 mv than that at C GC =0. A plausible explanation is that some GC molecules remain at the surface even in the close-packed state or almost all GC molecules are eliminated from the surface to be located just below the surface monolayer. To verify the previous hypothesis, the isotherm measurements and morphological observations for single and binary monolayers were typically done at C GC =50 μm. As seen in Fig. 10, all π A isotherms are expanded to larger areas compared with isotherms without GC (Figs. 2 and 4). At the closed-packed state, the molecular area above 40 mn m 1 for monolayers are similar to those for C GC =0, which means that most GC molecules are squeezed out of surface upon compression. It is noted that the positive slopes of ΔV A isotherms of PSM at C GC =50 μm suddenly become negative slopes at a molecular area of ~0.81 nm 2 (π=~19 mn m 1 ), as indicated by a solid arrow. This dramatic negative change is possibly due to desorption of GC from the air/water interface, which occurred simultaneously with conformational changes of monolayers. It is supposed that the growth of LC domains of PSM excludes GC molecules from the interface because pure PSM has LE/LC phase transition at ~21 mn m 1 (Fig. 2). In addition, conditions under which formation of a subphase is possible (5 M NaCl) to stabilize GC monolayers (typical LE phases) due to the salting out effect are evident in Fig. S8. n the contrary, GC cannot form stable monolayers on the subphase with lower NaCl concentrations. Therefore, in the current system, GC molecules are considered to be desorbed from the air/water interface by compression. At a minimum point of convex upward for ΔV, where the molecular area is ~0.66 nm 2 at 24 mn m 1, the slope again changes positively. This change means that the GC molecules are dramatically desorbed between surface pressures of ~19 and ~24 mn m 1, which correspond to the molecular areas at start (~0.81 nm 2 ) and end (~0.66 nm 2 ) points of the negative slope upon compression. The GC monolayer has positive ΔV values of ~503 mv at the close-packed state (Fig. S8), and the adsorbed GC layer also indicates positive ΔV values (Fig. S1(B)). Thus, the reduction in ΔV value (negative slope) means that GC is desorbed into the bulk by compression so that the ΔV value of the desorbed GC is canceled because of electrostatic neutrality in the bulk. Similarly, the desorption timing of GC by other monolayers is also estimated by slope variations in ΔV A isotherms as positive slopes change to negative slopes during desorption of GC. ΔV A isotherms of DPC-containing monolayers (DPC and PSM/DPC (1/1)) show that the slope changes negatively (dashed arrows). That is, GC is desorbed into the bulk between ~17 and ~19 mn m 1 for DPC and between ~12 and ~15 mn m 1 for PSM/DPC (1/1). n the contrary, as for ΔV A isotherms of CL-containing monolayers (CL and PSM/CL (1/1)), GC is found to be desorbed between ~10 and 19 mn m 1 for CL and between ~16 and 20 mn m 1 for PSM/ CL (1/1). owever, the ΔV A isotherms indicate no negative variations although the slopes change at surface pressures where GC begins to desorb from the surface, as opposed to other systems. Judging from

10 1280 S. Sakamoto et al. / Biochimica et Biophysica Acta 1828 (2013) Fig. 9. The π A and ΔV A isotherms of the ternary PSM/DPC/CL (1/1/1, by mol) monolayers on 0.02 M Tris buffer containing 0.13 M NaCl (p 7.4) in the presence of different amounts of GC (C GC =1, 5, 10, 25, and 50 μm) at K. The inset at the upper-left represents enlarged ΔV A isotherms in the surface potential region of 400 ΔV (in mv) 500, while the inset at the lower-right exhibits π t isotherms for adsorption of GC from the subphases containing different C GC. Note that the area axis means mean molecular area of the PSM/DPC/CL mixtures, where molecular weight and amount of adsorbed GC are not considered. Fig. 10. The π A and ΔV A isotherms of one-component (pure PSM, DPC, and CL) and two-component (PSM/CL (1/1, mol/mol) and PSM/DPC (1/1, mol/mol)) monolayers on 0.02 M Tris buffer with 0.13 M NaCl (p 7.4) at C GC =50 μm at K. Both solid and dashed arrows represent the break point where the slope of ΔV A isotherms changes negatively. Note that the area axis means mean molecular area of PSM, DPC, CL, and their mixtures, where molecular weight and amount of adsorbed GC are not considered. the corresponding π A isotherms, GC should be excluded from the surface. Thus, we postulate that GC molecules are desorbed from the surface by compression and then form a surface-associated aggregate just below the CL monolayer. This explanation may account for the lack of reduction in ΔV values and is supported by the favorable interaction between CL and GC, which can be induced by structural resemblance of the sterol skeleton between them (Fig. 1). Morphological analyses for single and binary monolayers at C GC =50 μm alsorevealtheimportance of CL in the interaction with GC (Fig. S9). Recently, complexation of CL and β-gc has been investigated by the nuclear magnetic resonance relaxation [73], which the authors used to show that stoichiometry of stable complexes between β-gc and CL is 2:1. This result raises the possibility that two molecules of GC bind to one molecule of CL at the air/water interface. Considering the previously discussed data, isotherm results in Fig. 9 suggest that a large number of GC molecules are gradually desorbed upon compression, and the ternary monolayer is strongly influenced by the adsorbed GC. After its exclusion from the surface, GC remains just below the CL monolayer through the attractive interactions between CL and GC. Fig. 11 shows the morphological images of the ternary monolayer as a function of π and C GC. Interestingly, FM and its corresponding BAM images demonstrate that raft domains (dark contrast in FM images) reduce in size, at which LC percentage is maintained constant irrespective of an increase in C GC. That is, the CL-rich domain is finely dispersed by surface-associated GC molecules. GC has been reported to lower membrane fluidity, which results in preventing infection from various types of viruses [43].urfindings denote that GC regulates the size of raft domains from underneath of monolayer to promote the LE network, through which transmitter substances are considered to be transported. Namely, this mechanism may control the fluidity in a cell membrane. To obtain more insight into the expanse of LE networks, the compressibility (Cs 1 )valueswerecalculatedfromπ A isotherms (Fig. S10). The results suggest that these values decrease when the amount of GC is increased, which means that the formation of LE network is promoted. owever, at a surface pressure of ~35 mn m 1, which corresponds to biomembrane pressure, morphological images show that the distinctive striped GC regions are clearly formed depending on C GC (obviously observed at a concentration of 50 μm), which means that GC divides the ternary monolayer at the interface into pieces. This division by GC is possibly due to a lack of CL at the surface relative to the amount of GC, even though two molecules of GC are estimated to bind to one molecule of CL [73]. Considering the characteristics of saponins, which exhibit membrane-permeabilizing activity, the striped regions derived from GC may most probably reflect membrane-disrupting activities, the so-called hemolytic activity of saponins [74]. Thus, the GC retained below the surface plays essential roles in the regulation of the raft domain and surrounding fluid networks. 4. Conclusions In this study, the surface behavior of GC with a lipid raft model consisting of PSM, DPC, and CL was systematically investigated by using the Langmuir monolayer technique. The π A and ΔV A isotherms and morphological observations reveal that GC regulates the size of raft

11 S. Sakamoto et al. / Biochimica et Biophysica Acta 1828 (2013) mn m mn m mn m mn m -1 PSM/DPC/CL (1/1/1) 39% 47% 47% 61% C GC = 1 µm 38% 43% 48% 61% C GC = 5 µm 39% 41% 51% 65% C GC = 10 µm 37% 41% 55% 64% C GC = 25 µm 39% 44% 54% 87% 50 µm C GC = 50 µm 100 µm Fig. 11. Fluorescence micrographs of the ternary PSM/DPC/CL (1/1/1, by mol) monolayers at four surface pressures (5, 15, 25, and 35 mn m 1 ) on 0.02 M Tris buffer with 0.13 M NaCl (p 7.4) at C GC =0, 1, 5, 10, 25, and 50 μm at K. The monolayers contain 1 mol% of the fluorescent probe (NBD-PC). The white and red scale bars in the lower-right represent 50 μm and 100 μm, respectively. The corresponding BAM images are shown in the upper-right of each photograph. The percentages indicate the ration of occupied LC domains in the frame. domains from underneath of monolayer to promote the formation of LE network, which plays an important role in the transport of substances via lipid rafts. Because a large amount of GC are found to be squeezed out of the surface upon lateral compression, the retaining GC just below CL monolayer plays a key role in the regulation of the size of raft domains and surrounding fluidity networks. In addition, our study using FM and BAM successfully shows activity associated with saponins, i.e., membrane-disrupting activity, at a GC concentration of 50 μm in the subphase despite the degree of GC's hemolytic activity is relatively low. Results of our investigation of the interaction of GC with one- and two-component systems suggest that CL is an essential component for formation of distinctive striped regions, which indicate membrane disruption by GC. In general, good and bad effects of saponins represent a double-edged sword.ur findings based on the present monolayer study should allow estimates of effects of saponins' interfacial behavior on biological membranes.